Myeloma mortality trends in the United States (Excluding Puerto Rico), 1999–2021, and future projections using ARIMA modeling and machine learning

Abstract Myeloma, including multiple myeloma, is the most prevalent plasma cell neoplasm in the United States and represents a significant contributor to cancer-related morbidity and mortality. This study analyzes myeloma mortality trends and employs machine learning to forecast future patterns. Mortality data for myeloma were obtained from the CDC WONDER cancer statistics database, spanning the years 1999 to 2021. Age-adjusted mortality rates (AAMRs) per 100,000 population were calculated using the 2000 U.S. Standard Population. Data were stratified by year, race, ethnicity, and gender. Join point regression analysis was used to estimate annual percent changes (APC) in AAMR with 95% confidence intervals (CIs). Forecasting was conducted using an ARIMA (Autoregressive Integrated Moving Average) model developed in Python 3 and run in Google Colab. A total of 262,339 deaths were attributed to myeloma during the study period. The overall AAMR was 2.7 for females and 4.3 for males. Among Hispanics, the average AAMR was 2.4 for females and 3.5 for males. In non-Hispanic populations, it was 2.8 and 4.4, respectively. The average AAMR for African American males was 7.8 compared to 5.5 for females. Non-Hispanic White individuals had an average AAMR of 4.1 for males and 2.5 for females. American Indians had an average AAMR of 2.9 for males and 2.1 for females. Asians had an average AAMR of 2.0 for males and 1.4 for females. Trends over time revealed an overall decline in AAMR across all demographic groups. Among Hispanic females, the AAMR decreased from 3.0 in 1999 to 2.0 in 2021 (APC -1.10, p<0.05). Non-Hispanic females AAMR decreased from 3.3 to 2.4 (APC -1.54, p<0.05). Hispanic males saw a modest decrease, from 3.4 to 3.0 (APC -0.79, p<0.05), and for non-Hispanic males the AAMR decreased from 4.8 to 3.9 (APC -0.96, p<0.05). Among African Americans, the AAMR for females decreased from 6.6 to 4.7 (APC -1.32, p<0.05) and from 8.8 to 7.1 (APC -0.89, p<0.05) for males. Among non-Hispanic White individuals, the AAMR for females decreased from 2.9 to 2.1 (APC -1.10, p<0.05) and from 4.4 to 3.6 (-0.82, p<0.05) for males. Among American Indians, the AAMR decreased from 2.9 to 1.2 for females (APC -2.23, p<0.05) and 4.6 to 2.1 for males (APC -1.33, p<0.05). Among Asians, the AAMR decreased from 2.0 to 1.1 for females (APC -1.51, p<0.05) and from 2.3 to 1.7 for males (APC -0.71, p<0.05). ARIMA-based forecasting was conducted using Python 3 via Google Colab to project AAMRs through 2031. The model suggested a general stabilization in trends from 2021 to 2031, with slight declines or plateauing across demographic groups. For African Americans, the predicted AAMR is 4.9 for females and 7.3 for males. For American Indians, the predicted AAMR for females is 1.9 and 3.1 for males in 2031. For Asians, the predicted AAMR is 1.2 for females and 1.9 for males. For non-Hispanic white females, it is predicted to be 1.8 compared to 3.6 for non-Hispanic males. For Hispanic females, the AAMR is predicted to be 2.2 and 3.5 for males. Overall, the model predicts an AAMR of 2.1 for females and 3.6 for males in 2031. Given the projected upward trends of mortality rates among African American males as well as Hispanic and American Indian males, further research and evidence-based, targeted interventions are warranted to address and reduce disparities in myeloma outcomes.